interaction term graph

[Daimin Roh]

hi
I want to draw a graph according to the value of the moderating variable in regression analysis.

- moderate variable is dummy, and independent also dummy
- t1~t4 is technology level(t1=1, othrewise=0; t2=1, otherwise=1; t3=1, otherwise=0; t4=1, otherwise=0)
- en is dummy variable and it is moderate variable


*code is here
reg y en t1 t2 t3 t4 c.t1#c.en c.t2#c.en c.t3#c.en c.t4#c.en if soreg_en=0
estimates store no
reg y en t1 t2 t3 t4 c.t1#c.en c.t2#c.en c.t3#c.en c.t4#c.en if soreg_en=1
estimates store yes
coefplot no yes, drop(_cons) xline(0)

but it is not working. and i am not sure it is correct. please help me

[Roman Mostazir]

Daimin, wellcome to the forum. The reason you are not getting reply is because you have not followed the forum posting rules. Have a look at the FAQ section before posting. Specially how to use -dataex- for providing data sample and how to use code delimiters to provide stata codes/results. Getting back to your question, since I have no idea of your data, I am guessing that you have single level data and by looking at your codes I don't feel you are doing it correctly. You should use the -margins- command to produce the interaction plot and your regression command should have the 'tXen' interaction term where 't' is the variable taking on values 1~4. You don't need several dummies for 't'. I provided a sample of fictitious data below to show how your data should be set up followed by codes of a regression model with interaction term and interaction plot:

clear

//Generate some fictitous data

set obs 100
g y = rnormal(20, 5)
g int t = (4-1+1)*runiform()+1
g en = runiform()>.5

//Data sample: Fictitous data

li y t en in 1/10, noobs clean

           y   t   en  
    30.12794   1    1  
    25.21316   2    0  
    21.48856   2    0  
    11.38934   4    0  
      16.354   3    0  
    24.30913   1    1  
    18.80323   3    1  
    22.58274   2    0  
    10.93992   4    1  
    14.92438   3    0

// Regression model:

 reg y i.t##i.en

      Source |       SS           df       MS      Number of obs   =       100
-------------+----------------------------------   F(7, 92)        =      1.59
       Model |  247.391535         7  35.3416479   Prob > F        =    0.1496
    Residual |  2051.18155        92  22.2954516   R-squared       =    0.1076
-------------+----------------------------------   Adj R-squared   =    0.0397
       Total |  2298.57308        99  23.2179099   Root MSE        =    4.7218

------------------------------------------------------------------------------
           y | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
           t |
          2  |   2.521532   1.927669     1.31   0.194    -1.306985     6.35005
          3  |   1.727287   1.828748     0.94   0.347    -1.904764    5.359337
          4  |    .254079   1.927669     0.13   0.895    -3.574438    4.082596
             |
        1.en |   4.146112   1.890235     2.19   0.031     .3919417    7.900282
             |
        t#en |
        2 1  |  -5.096853   2.630077    -1.94   0.056    -10.32041    .1267076
        3 1  |  -3.585629   2.745291    -1.31   0.195    -9.038015    1.866756
        4 1  |  -5.675484   2.699796    -2.10   0.038    -11.03751    -.313456
             |
       _cons |   17.85751   1.363068    13.10   0.000     15.15034    20.56468
------------------------------------------------------------------------------

//Means for interaction terms;

 margins t#en,

------------------------------------------------------------------------------
             |            Delta-method
             |     Margin   std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
        t#en |
        1 0  |   17.85751   1.363068    13.10   0.000     15.15034    20.56468
        1 1  |   22.00362   1.309593    16.80   0.000     19.40266    24.60459
        2 0  |   20.37904   1.363068    14.95   0.000     17.67187    23.08621
        2 1  |    19.4283   1.219165    15.94   0.000     17.00693    21.84967
        3 0  |    19.5848   1.219165    16.06   0.000     17.16343    22.00616
        3 1  |   20.14528   1.573935    12.80   0.000     17.01931    23.27125
        4 0  |   18.11159   1.363068    13.29   0.000     15.40442    20.81876
        4 1  |   16.58222   1.363068    12.17   0.000     13.87505    19.28939
------------------------------------------------------------------------------

//Generate plot:

marginsplot

[Daimin Roh]

Thank you so much.
A friend said that in order to draw a margin's graph of a categorical variable, you have to write a command in the same way as above.
The analysis results are the same, but I needed confidence in the graph results.

I want to ask one more question.
Two lines in the graph overlap (en=0;en=1) and are not well identified. I want to separate these two lines, so I'm looking for a way. What kind of command should I write?

[Andrew Musau]

You can look at mplotoffset from SSC or follow the following thread: https://www.statalist.org/forums/for...et-for-clarity

[Daimin Roh]

Cool! I drew the picture I wanted. Thank you.